Two of my best friends at CMU will defense next week

Cristian Dima
Active Learning for Outdoor Perception
Jun 28 2005, 2:00 PM, NSH 1507
Abstract
Many current state-of-the-art outdoor robots have perception systems that are primarily hand-tuned, which makes them difficult to adapt to new tasks and environments. Machine learning offers a powerful solution to this problem. Assuming that training data describing the desired output of the system is available, many supervised learning algorithms exist for automatically adjusting the parameters of possibly complex perception systems. This approach has been successfully demonstrated in many areas, and is gaining significant momentum in the field of robotic perception. An important difficulty in using machine learning techniques for large scale robotics problems comes from the fact that most algorithms require labeled data for training. Large data sets occur naturally in outdoor robotics applications, and labeling is most often an expensive process. This makes the direct application of learning techniques to realistic perception problems in our domain impractical. This thesis proposes to address the data labeling problem by analyzing the unlabeled data and automatically selecting for labeling only those examples that are hopefully important for the classification problem of interest. We present solutions for adapting several active learning techniques to the specific constraints that characterize outdoor perception, such as the need to learn from data sets with severely unbalanced class priors. We demonstrate that our solutions result in important reductions in the amount of data labeling required by presenting results from a large amount of experiments performed using real-world data.


Carl Wellington
Learning a Terrain Model for Autonomous Navigation in Rough Terrain
Jun 29 2005, 3:00 PM, NSH 1507
Abstract
Current approaches to local rough-terrain navigation are limited by their ability to build a terrain model from sensor data. Available sensors make very indirect measurements of quantities of interest such as the supporting ground surface and the location of obstacles. This is especially true in domains where vegetation may hide the ground surface or partially obscure obstacles. This thesis presents two related approaches for automatically learning how to use sensor data to build a local terrain model that includes the height of the supporting ground surface and the location of obstacles in challenging rough-terrain environments that include vegetation. The first approach uses an online learning method that directly learns the mapping between sensor data and ground height through experience with the world. The system can be trained by simply driving through representative areas. The second approach includes a terrain model that encodes structure in the world such as ground smoothness, class continuity, and similarity in vegetation height. This structure helps constrain the problem to better handle dense vegetation. Results from an autonomous tractor show that the mapping from sensor data to a terrain model can be automatically learned, and that exploiting structure in the environment improves ground height estimates in vegetation.